abstract

AAAI2021/tex/introduction.tex

@@ -50,7 +50,7 @@ the emergence of language and even intelligence among our pre-human ancestors.
 
 
 Yet, few works investigate the emergence of symbolic language with high
-compositionality in \emph{naturally} (i.e., without \emph{deliberately
+compositionality \emph{naturally} (i.e., without \emph{deliberately
   handcrafted} inductions). As a results, it is never clear whether \emph{natural}
 environment and agent are sufficient for compositionality. 
 In this work, we focus on generating high compositional symbolic language
@@ -58,9 +58,23 @@ naturally without any ``human'' induction.
 Initially, we thoroughly analyze the compositionality of emerged symbolic
 language after removing the \emph{deliberately handcrafted}
 inductions. Figure~\ref{fig:comp} reports the compositionality when train two
-agents in a listener-speaker referential game. It can be observed that \note{xxxxxxxx}.
-Thus, it is challenging to achieve high compositionality without induction.
-
+agents in a listener-speaker referential game. It can be observed that \note{it
+  is challenging to achieve high compositionality without induction as xxxxxx}. Moreover,
+we observe that the agent capacity plays a key role in compositionality, see
+Figure xxx.
+
+
+In this work, we reveal and characterize the quantitative relationship
+between the agent capacity and the compositionality of symbolic language both
+theoretically and experimentally.
+%theoretically
+Regarding the theoretical analysis, it is conduct based on the MSC
+(Markov Series Channel) model for the language transmission process and a
+novel mutual information-based metric for the compositionality.
+
+%experimentally
+Regarding the experimental verification, it is conducted on a listener-speaker referential game framework
+with eliminated unnatural inductions.
 
 ================